Exploration and augmentation of pharmacological space via adversarial auto-encoder model for facilitating kinase-centric drug development

Abstract

Predicting compound–protein interactions (CPIs) is of great importance for drug discovery and repositioning, yet still challenging mainly due to the sparse nature of CPI matrixes, resulting in poor generalization performance. Hence, unlike typical CPI prediction models focused on representation learning or model selection, we propose a deep neural network-based strategy, PCM-AAE, that re-explores and augments the pharmacological space of kinase inhibitors by introducing the adversarial auto-encoder model (AAE) to improve the generalization of the prediction model. To complete the data space, we constructed Ensemble of PCM-AAE (EPA), an ensemble model that quickly and accurately yields quantitative predictions of binding affinity between any human kinase and inhibitor. In rigorous internal validation, EPA showed excellent performance, consistently outperforming the model trained with the imbalanced set, especially for targets with relatively fewer training data points. Improved prediction accuracy of EPA for external datasets enhances its generalization ability, making it possible to gracefully handle previously unseen kinases and inhibitors. EPA showed promising potential when directly applied to virtual screening and off-target prediction, exhibiting its practicality in hit prediction. Our strategy is expected to facilitate kinase-centric drug development, as well as to solve more challenging prediction problems with insufficient data points.

Improving Compound–Protein Interaction Prediction by Self-Training with Augmenting Negative Samples

“…Our models outperformed the Random Undersampling , Weighted Loss , and Similarity-Controlled methods (Table ). We additionally implemented three types of 5-fold cross-validation recently adopted in CPI prediction and other fields: , Compound cross-validation (CV), protein CV, and compound–protein CV (see Supporting Information Text S2). As shown in Figure S2, our approach performed best in all the CVs, validating our self-training method’s effectiveness.…”

Off-targetP ML: an open source machine learning framework for off-target panel safety assessment of small molecules

“…Methods based on this approach are described as Proteochemometric modelling (PCM) [ 43 ]. Other studies have proposed pharmacological space augmentation via PCM and autoencoder models [ 44 ].…”

Molecular Generator for Multi-objective Optimization Based on the Pareto Algorithm

“…The models based on VAE reduct dimensionality to learn latent representations from input data .Therefore they can learn the probability distribution of the dataset and Reproduce the input data as much as possible. [5] CVAE, as a variant of VAE, introduces the concept of conditional vectors after transforming the input data into potential vectors. After decoding by the decoder, compound molecules with target ideal properties can be generated.…”